1. Filed of the Invention
The present invention relates to a video signal recording apparatus for recording a video signal in such a manner that the video signal is compressed before it is recorded.
2. Related Background Art
Hitherto, an Adaptive Discrete Cosine Transform, (hereinafter abbreviated to "ADCT") method, for example, disclosed as in Document 1 (ISO/JTC1/SC2/WG8N800) has been known as an image compression method. According to the above-described method, a function is realized which is capable of, regardless of the image content, compressing an image signal into a predetermined amount of encoded data by a repetition process in which plural trial and error operations are performed. The adjustment function thus-realized will now be described.
According to the above-described method, the amount of data is, as disclosed in Document 1, converged by using a Newton Raphson Method or a method disclosed in Document 2 ("Method of Controlling Code Amount in DCT Coding", 1989, Autumn National Conference of the Electronic Information Society, p. 45, Exposition No. D45, Pre-Exposition Theses, disclosed by a group including Nemoto). Furthermore, the data amount has been desired to be converged at an even higher speed. For example, according to Document 2, two or three repetitions will cause an effect to be obtained in that the data amount can be compressed to a desired ratio while revealing an effect of reduction ratio of errors which is smaller than .+-.5%.
According to the ADCT method, the compression ratio (that is, the amount of encoded data) can be controlled by using parameter F for quantizing image data. FIGS. 3A and 3B illustrate the relationship between quantization parameter F and amounts of encoded data. As shown in FIG. 3, the compression ratio is a monotone decrease function of F. FIGS. 3A and 3B illustrate the relationship between F of various images and compression ratios. As can be seen from FIGS. 3A and 3B, although the relationship between F and the compression ratio depends upon the content of the image, the compression ratio necessarily becomes a monotone decrease function. Therefore, by performing plural trial and error operations while adjusting F, image data can be converged to a desired compression ratio (amount of encoded data).
As an example of hardware capable of performing the high speed DCT operations required to transform the frequency in accordance with the above-described ADCT method, A121 manufactured by INMOS or the like has been available. The IC employed in A121 is capable of executing the DCT operation for one frame at about 30 milliseconds. Assuming that operations (for example, zigzag scanning and Huffman coding) except for the DCT can be subjected to a parallel process, it takes 30 milliseconds or longer to perform one trial-and-error operation (that is, the compression ratio is obtained by performing the compression while giving F of a certain value). In a case where an image taken by the CCD of an electronic still camera is compressed by the ADCT method before it is recorded to a magnetic disk, a time about 40 milliseconds is actually required because the time to perform recording to the magnetic disk is furthermore taken.
On the other hand, if the amount of encoded data may be changed depending upon the image in a case where the ADCT method is employed, a method may be employed which is arranged in such a manner that the conversion of the data amount by the repetition process is not performed but the compression is performed while fixing F. In this case, recording of about 25 images per second can be enabled. As a result, satisfactory continuous photographing performance for an electronic still camera can be realized. However, a problem arises in that the number of images recordable in a memory having a certain capacity becomes indefinite because the amount of encoded data undesirably becomes variable in length.
In order to record a predetermined number of images, the amount of encoded data must be converged by plural repetition operations as described above. Assuming that three times of trial-and-error operations must be performed to converge the amount of encoded data, only about 8 images can be recorded in one second because a time of about 120 milliseconds is taken. Therefore, satisfactory continuous photographing performance for an electronic still camera cannot be realized as yet.